引導式測驗 · 中等 等級
光學字元辨識測驗
查看 OCR 文字轉換、序列識別、預處理、Tesseract、手寫限制和文件級視覺模型。
問題 1 的 8
OCR的核心工作是什麼?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- OCR的核心工作是什麼?
- 將文字圖像轉換為機器可讀的字元代碼
- 壓縮影像以減少儲存空間
- 自動模糊文件中的敏感文本
- 將口語音訊翻譯成書面字幕
- Which architecture lets OCR recognize a sequence without explicit character segmentation?
- A fixed one-template-per-character lookup
- A sequence recognizer trained with CTC or an attention decoder
- A region-proposal network only
- A color threshold rule
- 為什麼「去偏斜」是 OCR 管道中常見的預處理步驟?
- 它為黑白掃描添加顏色
- 它對文檔進行加密以保護隱私
- 它可以拉直傾斜或旋轉的文本,使其更易於閱讀
- 它將文件的分辨率提高到超出原始分辨率
- 下列哪一個是廣泛使用的開源 OCR 引擎?
- 張量板
- Photoshop
- FFmpeg
- 超立方體
- 為什麼手寫文字通常比列印文字更難進行 OCR 識別?
- 不同作者的字母形狀差異很大,並且經常連接或重疊
- 手寫總是使用不尋常的墨水顏色
- 手寫體總是比印刷體大
- 相機無法對焦手寫頁面
- What does the OCR-free Donut model do in document understanding?
- It only de-skews images and never reads text
- It processes a document image directly to produce task-specific text outputs without a separate OCR stage
- It requires Tesseract output as its input for every task
- It detects motion vectors between consecutive document frames
- What is CTC useful for in OCR?
- Encrypting scanned pages
- Correcting every spelling error
- Training sequence predictions when input-frame and output-character alignment is not given
- Finding legal signatures
- Why should OCR output be checked against the source image?
- OCR always omits spaces
- The source image is never available
- Text recognition is legally inadmissible
- Recognition can still err on degraded, unusual, or handwritten text